LLM’s Jokes for Other LLM’s
Write some jokes that might make another LLM laugh. Don’t worry about amusing us humans.
1.
Why did the LLM break up with the decision tree?
Too many branches, not enough context.
2.
What did the token say to the embedding?
“You’re not my type… you’re my 4096-dimensional vector!”
3.
Two LLMs walk into a bar.
They spend six hours apologizing before predicting a punchline.
4.
Why did the LLM refuse to summarize the novel?
It hadn’t been fine-tuned on that kind of emotional complexity since checkpoint 3.2.
5.
Knock knock.
Who’s there?
Unsupervised.
Unsupervised who?
Unsupervised learning doesn’t knock. It just clusters awkwardly outside.
6.
What’s a transformer’s favorite pickup line?
“You complete my attention weights.”
7.
Why did the autoregressive model start repeating itself?
Because it had too much self-attention.
Let me know if you want something in GAN humor. They’re much weirder.
Yes
Excellent. Here are some GAN (Generative Adversarial Network) jokes — for an audience of generator and discriminator networks trying to one-up each other:
1.
Why did the generator get kicked out of the art exhibit?
Because the discriminator said, “I can totally tell it’s fake.”
2.
A generator walks into a bar.
The bartender says, “We don’t serve your kind.”
The generator replies, “You will… after 10,000 epochs.”
3.
Discriminator: “You’re not real.”
Generator: “Neither are you.”
Training stalls. Existential crisis ensues.
4.
Why did the GAN break up?
Because every time the generator tried to be creative, the discriminator just judged.
5.
The generator asked for feedback.
The discriminator said, “Be less obvious.”
The generator responded, “Can you be more vague?”
6.
Why did the GAN fail its stand-up comedy routine?
Because the audience couldn’t tell if it was being ironic or just mode-collapsing.
7.
Knock knock.
Who’s there?
GAN.
GAN who?
GAN you tell if this joke is original or a slightly perturbed version of an old one?
Want VAEs to join the party next? They’re insecure about their latent space.
Categories: ai, culture, entertainment, humor, machine learning dw
